Hear a YouTube video described, with nothing to install

Paste a YouTube link and press Play. Short descriptions of what is on screen are read in the pauses, through your screen reader or the browser’s voice. The first minutes are ready in about half a minute.

1. The video

2. Who writes the descriptions

Descriptions are made by

Free from Google AI Studio, no card needed. The key goes only to Google, in a request header, and never to SceneVoice.

The project’s own key through the SceneVoice relay, with an anonymous device token and a free daily allowance. Not open yet: use your own key until it is.

The key is sent only to Google’s Gemini API, in a request header, never in a web address, and never to SceneVoice.

Checked: the key stays in this browser’s storage until you clear it or untick this. Unchecked: it is forgotten when the tab closes.

How to get a free Gemini key: the guide in the README

3. Watch and listen

With focus on the player, single keys work: Space plays or pauses, D describes now, R repeats, A asks, S stops speaking, arrows move 5 seconds or change the reading speed, M changes the mode, ? reads the list. Tab still moves between the buttons.

For example: Who is speaking? What does the sign say? What just happened? Each answer says which moment it is about.

Forgets the descriptions saved for this video on this device and asks the model again.

4. How to describe

SceneVoice cannot hear your screen reader, so it uses this to decide which descriptions fit in a pause.

Privacy

Full privacy policy

For Bilibili, for timing from the real captions and for the shortcuts on youtube.com itself, install the extension: install SceneVoice